[1]楚浩宇,高 萌,刘永生.基于并行组合分类器的脱机手写体数字识别[J].计算机技术与发展,2018,28(03):105-108.[doi:10.3969/ j. issn.1673-629X.2018.03.022]
　CHU Hao-yu,GAO Meng,LIU Yong-sheng.Off-line Handwritten Digit Recognition Based on Parallel Combined Classifiers[J].,2018,28(03):105-108.[doi:10.3969/ j. issn.1673-629X.2018.03.022]
点击复制

基于并行组合分类器的脱机手写体数字识别()
分享到：

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
28
期数:
2018年03期
页码:
105-108
栏目:
智能、算法、系统工程
出版日期:
2018-03-10

文章信息/Info

Title:
Off-line Handwritten Digit Recognition Based on Parallel Combined Classifiers
文章编号:
1673-629X(2018)03-0105-04
作者:
楚浩宇; 高 萌; 刘永生
东北农业大学 电气与信息学院,黑龙江 哈尔滨 150030
Author(s):
CHU Hao-yu; GAO Meng; LIU Yong-sheng
School of Electrical and Information,Northeast Agricultural University,Harbin 150030,China
关键词:
模式识别; 组合分类器; LR; 广义回归神经网络; 支持向量机; 手写体数字
Keywords:
pattern recognition; combined classifiers; LR; GRNN; SVM; handwritten digit
分类号:
TP391.4
DOI:
10.3969/ j. issn.1673-629X.2018.03.022
文献标志码:
A
摘要:
为了提高脱机手写体数字识别的识别率和可靠性,并且考虑到传统的单一分类器对数字之间差异的敏感性不同,综合 K-近邻算法、广义回归神经网络、支持向量机三种机器学习算法,提出了一种并行组织结构的组合分类器。 并行组合分类器通过改进的投票机制来判定识别结果。 以 MNIST 数据库为数据来源,在 MATLAB 平台上开展各种分类器的性能对比实验。 组合后的识别率、拒识率、误识率、可靠性分别可达到 97. 48％、1. 55％、0. 97％、99. 02％。 实验结果表明,并行组合分类器在鲁棒性方面优于传统的单一分类器,在识别率、拒识率、算法的时间复杂度上均优于其他组合分类器。 并行组合分类器以简易结构实现了脱机手写体数字的快速、高效识别。
Abstract:
In order to improve the recognition rate and reliability of off-line handwritten digit recognition,considering that traditional single classifiers have different sensitivity to the differences between digital,we propose a combined classifier of parallel organizational struc-ture combining three machine learning algorithms of K-nearest neighbor,general regression neural network and support vector machine.It uses the improved voting mechanism to determine the recognition result. Using MNIST database as data source,the comparable experi-
ment on the performance of classifiers is carried out on MATLAB,whose results (recognition rate,rejection rate,false accept rate,reliability) are 97. 48％,1. 55％,0. 97％ and 99.02％. The experiments indicate that the parallel combined classifier is superior to the traditional single classifier in terms of robustness,and other combined classifiers in terms of recognition rate,rejection rate and time complexity. With a simple structure,it can achieve fast and efficient off-line handwritten digit recognition.

相似文献/References:

[1]鲁晓南 接标.一种基于个性化邮件特征的反垃圾邮件系统[J].计算机技术与发展,2009,(08):155.
　LU Xiao-nan,JIE Biao.An Individual Anti- Spam Technology[J].,2009,(03):155.
[2]杨盈 李朝峰.基于距离学习法的模式识别[J].计算机技术与发展,2009,(03):77.
　YANG Ying,LI Chao-feng.Learning Using Distance Based Training Algorithm for Pattern Recognition[J].,2009,(03):77.
[3]张充 史青宣 苗秀芬 杨芳 田学东.基于BP神经网络的手写体数字识别[J].计算机技术与发展,2008,(06):128.
　ZHANG Chong,SHI Qing-xuan,MIAO Xiu-fen,et al.Handwritten Numeral Recognition Based on BP Neural Network[J].,2008,(03):128.
[4]万忠 张铃 张燕平 陈洁.基于覆盖算法的两层结构分类器设计[J].计算机技术与发展,2007,(01):65.
　WAN Zhong,ZHANG Ling,ZHANG Yah-ping,et al.A Two- Layer Classifier Design Method Based on Covering Algorithm[J].,2007,(03):65.
[5]钟淑瑛 李陶深.基于MATLAB的BP—LVQ神经网络组合分类模型[J].计算机技术与发展,2006,(02):114.
　ZHONG Shu-ying,LI Tao-shen.BP- LVQ Neural Networks Classifiers Combination Model Based on MATLAB[J].,2006,(03):114.
[6]宋丹 徐蔚鸿.基于模糊理论的车型识别[J].计算机技术与发展,2006,(03):47.
　SONG Dan,XU Wei-hong.Recognition of Automobile Type Based on Fuzzy Theoretics[J].,2006,(03):47.
[7]陈晓艳 董朝轶 李永亭 刘月文.基于多层感知器网络的农作物疾病诊断系统[J].计算机技术与发展,2011,(11):84.
　CHEN Xiao-yan,DONG Chao-yi,LI Yong-ting,et al.Study of Diagnosis for Diseases of Agricultural Crops Based on Multi-Layer Perceptrom Networks[J].,2011,(03):84.
[8]汪晨 张涛 林为民 邓松 时坚 李伟伟.图像识别综述及在电力信息安全中的应用研究[J].计算机技术与发展,2012,(04):161.
　WANG Chen,ZHANG Tao,LIN Wei-min,et al.Image Recognition Review and Application Research in Electric Power Information Security[J].,2012,(03):161.
[9]陈静,邱晓晖,孙娜. 基于二维Gabor小波与SPP算法的人脸识别研究[J].计算机技术与发展,2014,24(11):110.
　CHEN Jing,QIU Xiao-hui,SUN Na. Research on Face Recognition Based on 2 D Gabor Wavelet and SPP Algorithm[J].,2014,24(03):110.
[10]檀何凤,刘政怡. 模式分类方法比较研究[J].计算机技术与发展,2015,25(02):99.
　TAN He-feng,LIU Zheng-yi. Research on Comparison of Pattern Classification Method[J].,2015,25(03):99.

更新日期/Last Update: 2018-04-26